Mathematical characterisation and performance analysis is a key step in the design and development of wireless communication systems. It allows a system designer to study and compare the performance under a wide range of system parameter settings without computationally-intensive simulations. This, in turn, facilitates easier performance optimisation, which leads to enhanced quality of service for the end-users. In this work, we explore the utility of tools from extreme value theory (EVT) for the statistical characterisation of the higher-order statistics of signal quality metrics in wireless communication. All of the existing literature in this domain assumes that the order statistics are evaluated over sequences of independent and identically distributed random variables (RVs), whereas this assumption is restrictive in most wireless communication systems. We study how tools from EVT can be used to evaluate the higher-order statistics of sequences of independent and non-identically distributed signal to noise ratio RVs in an energy harvesting relay assisted communication system. We also demonstrate how these results can be used to arrive at meaningful inferences regarding the system performance which are otherwise analytically intractable.